Beez content methodology
Current Beez launch disclosure and the target research, sourcing, human review, calculation, translation, and update methods.
Research question and scope
Each page starts with one reader question, a defined search intent, a jurisdiction, and a scope boundary. The boundary prevents a broad guide and its supporting articles from repeating the same answer. It also records what the page cannot determine, such as personal suitability, legal entitlement, product eligibility, or an individual Sharia ruling.
Source hierarchy
For UAE and GCC legal, regulatory, statistical, and financial-system claims, we begin with the authority responsible for the rule or data. Preferred sources include government portals, ministries, regulators, central banks, official rulebooks, exchanges, statistical agencies, and recognised Islamic-finance standards bodies.
Every source is reopened during final review. Reviewers check jurisdiction, effective date, population or worker category, definitions, amendments, and whether a document describes a current rule or a historical period. Historical data is labelled with its period and is never presented as live.
Claim-level citations
A citation is attached to the sentence or block it supports. The content record stores a source identifier, title, publisher, URL, access date, effective date when relevant, jurisdiction, and claim context. A link alone is not treated as verification.
Original contribution
A public page must do more than restate its sources. Depending on the topic, the original contribution may be a plain-language decision tree, comparison matrix, checklist, annotated document, local hypothetical scenario, transparent formula, or connection to a Beez lesson or private tool. Examples are labelled hypothetical and do not use reader data.
Review requirements
The CMS derives mandatory review stages from the document type, disclaimer class, source language, and requested specialist controls. Decisions are recorded against the exact revision checksum. The publishing system rejects missing, stale, unauthorised, rejected, or self-approved decisions.
Financial education receives financial review. Islamic-finance interpretation receives both financial and qualified Sharia review. Translated content receives locale review. Legal and regulatory claims are checked by an appropriately qualified reviewer for the stated jurisdiction. Final publication requires a named human publisher.
Calculators
Each calculator publishes its formula, assumptions, limitations, and a worked hypothetical example. Formula and boundary tests are checked before release. Values entered by a reader stay local to the browser and are excluded from URLs, logs, analytics, metadata, and unmasked replay.
Calculator outputs are estimates. The surrounding explanation identifies cases where an official authority, contract, employer, financial institution, or qualified adviser must make the actual determination.
Lessons and video
A curated lesson page requires a verified video source, duration, upload date, caption or transcript, accessible poster text, media rights, and a knowledge check. Academy inventory is selected deliberately; it is not automatically converted into thousands of public pages.
Translation
Translations belong to a shared translation group but have their own locale, reviewer, dates, and publication state. A machine translation may be used as a draft only. A locale reviewer checks meaning, terminology, links, examples, disclaimers, and local relevance before publication.
AI-assisted work
AI assistance is recorded in content provenance. It may help organise sources or draft language, but it does not count as source verification, specialist review, or publication approval. Human reviewers check every claim and retain responsibility for the final page.
Freshness and versioning
Time-sensitive pages display a meaningful as-of date and a future review date based on their risk and update burden. Event-driven reviews follow material changes to a law, standard, official process, dataset, source, or formula. Editing a page creates a new revision and invalidates prior approvals.
Published content is exported deterministically from the LMSService database into the Next.js build. Only approved published documents enter raw HTML, canonical metadata, feeds, and sitemaps. Preview and fixture content remain blocked from indexing.